电力系统状态估计并行算法实现

发布时间：2018-01-24 09:33

本文关键词： 电力系统状态估计 OpenMP技术节点分配法网络分割并行计算　出处：《郑州大学》2014年硕士论文　论文类型：学位论文

【摘要】：电力系统状态估计在现代电力系统中有着很重要的实用性意义，它是电力网络中电能管理系统（Energy Management System，EMS）关键性的组成部分。状态估计主要是电力系统网络实现在线调控，进行经济调度，达到安全评估，为电力网络系统的下一步决策提供精确实时的操作和运行数据，为电网的工作人员提供指导，因此它的实用性作用不容忽略，具有重要的研究价值。电力系统中状态估计理论基础和实践应用发展迅速，技术水平得到了大幅度的提高，然而，随着电网规模的不断扩大，这就让电网的各项指标数据在进行分析时需要处理的计算量也逐步增加，，很多问题也就显现出来。目前，电力系统状态估计的算法虽说在不断提高，但还存在一定的弊端，比如计算速度、内存占用以及收敛可靠性也不尽如人意。近几年来，随着计算机硬件的发展，使得并行计算的优势显露出来，因此，目前对电力系统的并行研究也越来越多。本文介绍了电力系统状态估计研究背景及意义，阐述状态估计目前的发展动态和应用的新技术，对状态估计的数学模型进行了详细描述，分析较为常用的经典的电力系统状态估计算法，对它们的原理和特点进行研究比较。采用并行的原理进行分析，利用OpenMP(Open Multi-Processing)编程模式，结合实际电网系统的特点，进行了基于OpenMP编程环境下的并行算法研究。算法的改进主要表现在两方面：一方面是通过对原有最小二乘法的串行算法结构进行分析，根据原始的串行程序的结构特点，利用OpenMP的数据并行化编译语句，对线性方程组矩阵的求和以及矩阵乘积的进行并行改写；另一方面对实际的网络特点进行分析，根据编译环境的运行特点，对网络进行分割，将整体系统划分为两个子系统，利用工作分区编译指令进行程序改写，实现粗粒度的模块化并行。此外，编程工作中也采用了一些提升并行性能的方法，尽量较小开销。本文在Visual Studio2008的平台和Intel Fortran11的编译环境中进行程序的改进，然后不同规模的网络算例进行试验，对基于OpenMP编译环境的改进后并行程序的数据结果进行分析，验证改进后并行算法的有效性。
[Abstract]:Power system state estimation is of great practical significance in modern power system. It is an energy Management System in power network. The key component of EMS. state estimation is to realize on-line regulation and control of power system network, to carry out economic dispatch and to achieve security evaluation. It provides accurate and real-time operation and operation data for the next decision of power network system, and provides guidance for the staff of power network, so its practicability can not be ignored. It has important research value. The theory foundation and practical application of state estimation in power system have developed rapidly, and the technical level has been greatly improved. However, with the continuous expansion of power grid scale. As a result, the amount of calculation needed to be dealt with in the analysis of the grid index data increases gradually, and many problems appear. At present, the algorithm of power system state estimation is improving. However, there are still some drawbacks, such as computing speed, memory usage and convergence reliability are not satisfactory. In recent years, with the development of computer hardware, the advantages of parallel computing are revealed. At present, there are more and more parallel researches on power system. This paper introduces the research background and significance of power system state estimation, describes the current development and application of state estimation, and describes the mathematical model of state estimation in detail. The classical power system state estimation algorithms are analyzed, and their principles and characteristics are studied and compared. The parallel principle is adopted, and the programming mode of OpenMP(Open Multi-Processingis used to combine the characteristics of the actual power grid system. The research of parallel algorithm based on OpenMP programming environment is carried out. The improvement of the algorithm is mainly in two aspects: on the one hand, through the analysis of the original structure of the serial algorithm of the least square method. According to the structural characteristics of the original serial program, the summation of linear equations matrix and the parallel rewriting of matrix product are carried out by using the data parallelization of OpenMP. On the other hand, the characteristics of the actual network are analyzed. According to the running characteristics of the compiling environment, the network is divided into two subsystems, and the program is rewritten by compiling instructions in the working area. The coarse-grained modularization parallelism is implemented. In addition, some methods to improve the parallel performance are used in the programming work to minimize the overhead. This paper improves the program in the platform of Visual Studio2008 and the compiling environment of Intel Fortran11, and then carries on the experiment with the network example of different scale. The data results of the improved parallel program based on OpenMP compile environment are analyzed to verify the effectiveness of the improved parallel algorithm.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM73

【参考文献】